Reliability of malaria microscopy in epidemiological studies: results of quality control

To assess the interrater reproducibility of malaria microscopy in epidemiol ogical studies, 711 thick blood films from population-based surveys were ra ndomly selected and reread by 4 experienced microscopists. Sample estimates of the prevalence of P. falciparum infection, geometric mean parasite dens ity and the proportion of samples above various parasite density cut-off le vels were almost identical in the routine and quality control readings. Dif ferences were, however, encountered in the sample estimates for gametocyte ratio, proportion of mixed infection and average density index. In all thre e cases the quality control result was significantly higher than the routin e evaluation. On the level of the individual slide there was good interrate r agreement for the presence of P. falciparum infections (Kappa index kappa = 0.79) which was even better when parasite densities between 4 and 100/mu l were excluded (kappa = 0.94). With respect to the assessment of parasite density, a high level of disagreement was found. While the mean difference between the two readings was not different from 0, the second reading was between 0.12 and 10 times that of the first. However, the level of disagree ment significantly fell with increasing parasite densities. Thus malaria mi croscopy is very reliable for the estimation of parasite ratios and geometr ic mean parasite densities within and between studies as long as the same m ethodology is used, but tends to underestimate the gametocyte ratio and pro portion of mixed infections. Care must be taken, however, when individual p arasite density is related to other explanatory variables, due to the high degree of variability in the parasite enumeration.